Interactive Stratospheric Aerosol Microphysics‐Chemistry Simulations of the 1991 Pinatubo Volcanic Aerosols With Newly Coupled Sectional Aerosol and Stratosphere‐Troposphere Chemistry Modules in the NASA GEOS Chemistry‐Climate Model (CCM)

We have coupled the GEOS‐Chem tropospheric‐stratospheric chemistry mechanism and the Community Aerosol and Radiation Model for Atmospheres (CARMA), a sectional aerosol microphysics module, within the NASA Goddard Earth Observing System Chemistry‐Climate Model (GEOS CCM) in order to simulate the interactions between stratospheric chemistry and aerosol microphysics. We use observations of the 1991 Mount Pinatubo volcanic cloud to evaluate this new version of the GEOS CCM. The GEOS‐Chem chemistry module is used to simulate the oxidation of sulfur dioxide (SO2) more realistically than assuming hydroxyl radical (OH) fields are constant, as OH concentrations in the plume decrease dramatically in the weeks following the eruption. CARMA simulates sulfate aerosols with dynamic microphysical and optical properties. The CARMA‐calculated aerosol surface area is coupled to the chemistry module from GEOS‐Chem for the calculation of heterogeneous chemistry. We use a set of observational and theoretical constraints for Pinatubo to evaluate the performance of this new version of the GEOS CCM. These simulations are specifically compared with satellite and in‐situ observations and provide insights into the connections between the gas‐phase chemistry and the aerosol microphysics of the early plume and how they impact the climatic and chemical changes following a large volcanic eruption. A second, smaller eruption is also included in these simulations, the 15 August 1991, eruption of Cerro Hudson in Chile, which we find essential in explaining the aerosol optical depth in the Southern Hemisphere in 1991.

In addition to surface cooling, the increased aerosol loading from volcanic eruptions leads to enhanced absorption of near-infrared and terrestrial longwave radiation and to a subsequent warming of stratospheric temperatures (Angell, 1997;Free & Lanzante, 2009). Circulation changes due to this heating, alongside changes to heterogeneous chemistry and photochemistry, cause changes to global ozone chemistry (Aquila et al., 2013;Bekki et al., 1993;Dhomse et al., 2015;Kinne et al., 1992;Prather, 1992;Solomon, 1999). Volcanic aerosols that reach polar latitudes also contribute to enhanced spring-time ozone depletion and have been associated with years of anomalously severe ozone holes (Hofmann & Oltmans, 1993;Solomon et al., 2016;Stone et al., 2017;Tabazadeh et al., 2002;Zhu et al., 2018). This ozone depletion has been identified in the years 1992 and 1993 as the Pinatubo aerosol reached high Southern latitudes (Hofmann & Oltmans, 1993;Kerr, 1993;Knight et al., 1998).
Despite the numerous studies on the 1991 eruption of Mt. Pinatubo (e.g., LeGrande et al., 2015;Marshall et al., 2022;Yang et al., 2019), questions remain about the parameters of the eruption and how to represent it in global models. For example, the estimated amount of the SO 2 , sulfate, and ash injected varies widely across these studies. Additionally, Pinatubo was unique in the satellite record for its spread from the Northern tropics to the Southern Hemisphere shortly after the eruption (Pitari et al., 2016). These prior modeling studies have emphasized the importance of injection altitude in correctly modeling the transport of the Pinatubo plume (Aquila et al., 2012;Timmreck et al., 1999), but observations define a large range of plausible altitudes for the eruption (Holasek et al., 1996;Tupper et al., 2005). One key driver in this uncertainty is sparse observations of the thick plume's vertical structure near the time of the eruption, in part caused by near-total attenuation of the sunlight observed by occultation measurements, as well as the confounding impacts of coincident meteorology, including Typhoon Yunya (Holasek et al., 1996). This leaves researchers to "tune" the injection height in simulations to compensate for a particular model's transport characteristics. Even so, the unique southward shift of the Pinatubo aerosol into the Southern Hemisphere is challenging to re-create with global models. Multiple mechanisms have been proposed to explain this: stratospheric dynamical changes due the radiative interaction of the volcanic sulfate (Dhomse et al., 2020;Young et al., 1994); other uncertainties associated with the injection parameters of Pinatubo (e.g., magnitude, timing, and initial horizontal spread) (Niemeier et al., 2009); and the omission of the 15 August 1991, Cerro Hudson eruption in most models .
Another uncertainty associated with modeling the Pinatubo plume in global models is whether the eruption altered the local oxidizing capacity of the atmosphere and how that impacted the conversion of SO 2 to aerosol. Early global chemistry modeling studies of El Chichon (e.g., McKeen et al., 1984) and Pinatubo (Bekki, 1995) assumed the loss of OH due to its reaction with SO 2 to be negligible for a Pinatubo-sized eruption. More recently, however, Mills et al. (2017) showed, using a model with interactive oxidants, that OH can decrease by as much as 95% within the sulfur dense environment of the Pinatubo plume, slowing down the oxidation of SO 2 to sulfuric acid and sulfate aerosol.
In this study, we revisit the Pinatubo eruption using an updated version of the NASA Goddard Earth Observing System (GEOS) Earth system model, now coupled with the aerosol sectional microphysics module Community Aerosol and Radiation Model for Atmospheres (CARMA) and the GEOS-Chem tropospheric and stratospheric chemistry mechanism. With respect to previous volcanic studies with the GEOS model (Aquila et al., 2012(Aquila et al., , 2013(Aquila et al., , 2021 this updated version of the GEOS CCM includes, for the first time, a representation of the evolving volcanic aerosol size distribution in CARMA and a comprehensive chemistry mechanism in GEOS-Chem. This is, to our knowledge, the first simulation of the Pinatubo plume performed with the stratosphere-troposphere GEOS-Chem mechanism. We refer to this combined model as the GEOS Chemistry-Climate Model (GEOS CCM).
Below we outline and evaluate the capabilities of this new model and use it to simulate the 1991 eruption of Pinatubo. The extensive literature around observing and modeling this eruption provide many ways to understand this novel model in the context of the observed eruption and prior modeling efforts. Secondarily, the ensemble of simulations can provide insight into the remaining uncertainties in the Pinatubo literature. We estimate the relative importance of transported Pinatubo aerosols and Cerro Hudson aerosols in Southern Hemisphere in 1991 and we explore the hypothesis that OH depletion slowed initial SO 2 oxidation rates in the Pinatubo plume.

Model Description
GEOS is an Earth system model based on the architecture of the Earth System Modeling Framework (Hill et al., 2004;Molod et al., 2015). In this study, we use the atmospheric general circulation model (AGCM) configuration in its "free-running" mode; the model calculates its own meteorology without any data assimilation and with imposed sea surface temperatures based on observations. The GEOS system has been shown to perform well in stratospheric chemistry and transport processes (Douglass et al., 2012;SPARC CCMVal, 2010;Strahan et al., 2011). We run GEOS (version Icarus 3.3) at a ∼100 km horizontal resolution on a cubed-sphere grid with 72 hybrid-sigma vertical levels extending from the surface to ∼80 km. While the GEOS AGCM can be coupled to various aerosol modules, here we are using the sectional aerosol microphysics from the Community Aerosol and Radiation Model for Atmospheres (CARMA, Bardeen et al., 2008;Toon et al., 1988). In this work we have coupled CARMA to the GEOS-Chem tropospheric and stratospheric chemistry mechanism (Bey et al., 2001; acmg.seas.harvard.edu/geos/), and both the CARMA aerosol and GEOS-Chem chemistry (simulated ozone, water vapor, methane, nitrous oxide,  are radiatively interactive within the model. This coupling of aerosols and chemistry modules within the GEOS framework is referred to as the GEOS Chemistry-Climate Model (GEOS CCM, Nielsen et al., 2017).
CARMA is a sectional aerosol microphysics model, originally developed as a 1-dimensional package (Turco et al., 1979), that can be configured to simulate aerosols of many different types; past studies have used it to simulate sulfates, volcanic ash, dust, carbonaceous aerosols, and nitrates. CARMA v3.0 has been used extensively in other global models (e.g., Bardeen et al., 2008;English et al., 2011). This version of CARMA was introduced in GEOS for application to dust (Colarco et al., 2014) and has subsequently been expanded to include sulfate aerosols (Chen et al., 2018). To fit the focus of this study, we have configured CARMA to simulate only sulfate aerosols. Sensitivity experiments have shown a sulfate simulation using 24 size bins, spread logarithmically in volume between 0.25 nm and 6.7 μm in radius reasonably represents the background stratospheric aerosol observed by in situ optical particle counter (OPC) observations. CARMA simulates the nucleation, condensational growth and evaporation, coagulation, and settling of sulfate aerosols within the GEOS-CCM model, following the mechanism of English et al. (2013). Other aerosol species are calculated using the bulk aerosol microphysics package as part of GEOS-Chem and are not directly interactive with CARMA.
The GEOS-Chem Chemistry Transport Model was originally an offline 3-D chemical transport model driven by assimilated meteorological data derived from the GEOS Earth system model and analysis system (Bey et al., 2001). GEOS-Chem (v10-1) has since been implemented interactively in the GEOS AGCM as a tropospheric and stratospheric chemistry mechanism (Hu et al., 2018;Keller et al., 2021). Note that Keller et al., 2021 uses a more recent version of the GEOS-Chem mechanism (v12.1) than the model described here (v10-1). GEOS-Chem's chemistry calculation is the most comprehensive chemistry mechanism in the GEOS framework and here we specifically take advantage of GEOS-Chem's interactive oxidants and sulfur cycle to simulate the chemistry of stratospheric volcanic plumes.
For this study, CARMA aerosol is coupled to the GEOS-Chem chemistry module via: the production and loss of sulfuric acid, both by chemical reactions in the chemistry mechanism and nucleation, condensation, and evaporation within CARMA; aqueous aerosol production; and heterogeneous chemistry. This coupling is shown diagrammatically in Figure 1. The GEOS-Chem chemistry mechanism tracks sulfuric acid/sulfate precursors (SO 2 , OCS, etc.) and produces both gaseous sulfuric acid as well as aqueously produced sulfate, which are both passed to CARMA. SO 2 oxidation by hydroxyl radical is the main oxidation pathway for volcanic SO 2 . In order to calculate the nucleation rate, condensation, and evaporation of sulfate aerosols, CARMA tracks the sulfuric acid vapor tracer. After its microphysics are calculated, CARMA provides GEOS-Chem with aerosol bulk mass and aerosol surface area for heterogeneous chemistry calculations and any loss and production of sulfuric acid vapor by condensation or evaporation on sulfate aerosols. This setup includes the stratosphere-troposphere chemistry from GEOS-Chem, the detailed aerosol microphysics calculations of CARMA, and the heterogeneous chemistry calculations within GEOS-Chem using CARMA derived aerosol properties, coupled within the GEOS framework. While the CARMA microphysics and the coupling of the CARMA aerosols to the radiation code use the CARMA aerosol properties, the current version of this model uses the CARMA total mass but assumes a lognormal aerosol size distribution for the calculation of effective radius and optical properties in GEOS-Chem's photolysis codes. This approach to photolysis will be updated to use CARMA optical properties in a future version of the model. The CARMA aerosols are fully radiatively interactive with the GEOS Earth system model, as are the GEOS-Chem gas phase species (simulated ozone, water vapor, methane, nitrous oxide, CFC-11, CFC-12, and HCFC-22).

Initial Conditions and Injection Parameters
The model was spun up at perpetual 1990 conditions of ozone depleting substances, greenhouse gases, and aerosols for 10 model years prior to conducting the experiment. Non-volcanic stratospheric sulfur comes primarily from biogenically generated surface emissions of OCS that is relatively inert in the troposphere but oxidizes to SO 2 in the stratosphere and then forms sulfate. Stratospheric sulfate results also from anthropogenic and natural SO 2 sources (and precursors; e.g., dimethyl sulfide [DMS]) in the troposphere that are transported to the stratosphere (Kremser et al., 2016). Anthropogenic emissions in the model come from the EDGAR HTAP database (Janssens-Maenhout et al., 2015). We have verified both the loading and simulated size distribution of the CARMA-calculated background stratospheric aerosol layer based on these emissions using balloon-borne OPC observations. Degassing (non-explosive) volcanoes also provide a source of SO 2 but are not included as a source in this study. A constant surface boundary condition of OCS of 500 pptv and a DMS emissions climatology, respectively, are prescribed in the model. We use Pinatubo injection parameters similar to those in Mills et al. (2016). For the primary ensemble, we inject the Pinatubo SO 2 uniformly between 0° and 15°N from 18 to 21 km at 120°E on 15 June 1991. The location of this injection is based on Total Ozone Mapping Spectrometer (TOMS) observations in the days following the eruption (Bluth et al., 1992). We use this distributed injection to account for the unique meteorological situation in the region at the time of the eruption, which caused rapid southward transport. To test the sensitivity to the injection amount and to account for the uncertainty associated with the Pinatubo injection parameters, we completed a three-member simulation ensemble injecting 10 Tg of SO 2 (5 Tg S) and a separate three-member ensemble injecting 20 Tg of SO 2 (10 Tg S).
Cerro Hudson was a smaller eruption than Pinatubo with less extreme local meteorology, so we use more localized injection parameters. Based on SO 2 mass estimates by Carn et al. (2016) using TOMS, HIRS/2, and SBUV data, we inject 2.6 Tg of SO 2 (1.3 Tg S) from 13 to 18 km directly over the volcano over 24 hr on the day of the eruption, 15 August 1991. Cerro Hudson has been included in both the 10 Tg and 20 Tg ensemble. To evaluate the individual impact of Pinatubo, we also completed a simulation without the Cerro Hudson, using the 20 Tg Pinatubo injection parameters.
We summarize the set of Pinatubo model experiments in Table 1.

Observational Datasets
In order to validate the model, we use various space-based sensors as well as balloon-borne OPC measurements. The main satellite instrument used to observe stratospheric aerosols during the Pinatubo era is the second Stratospheric Aerosol and Gas Experiment (SAGE II) flown on board the Earth Radiation Budget Satellite from 1984 to 2005. SAGE II made solar occultation measurements and provided a long-term space-based record of stratospheric aerosol, and alongside its predecessors and successors we have a record extending from the 1970s to the present. The version of the SAGE II record used in this study is part of the Global Space-based Stratospheric Aerosol Climatology (GloSSAC), which has supplemented the SAGE observations with available air-and ground-based LIDAR as well as other satellite observations (Kovilakam et al., 2020). These supplemental data are especially important in the months following the Pinatubo eruption during which time SAGE II was not able to make measurements because of the extreme optical thickness of the resulting stratospheric aerosol layer (Kovilakam et al., 2020).
In addition to SAGE II, we use an aerosol record of the Pinatubo era from the Advanced Very High-Resolution Radiometer (AVHRR) on the NOAA-11 satellite (Long & Stowe, 1994). This record is based on backscattered sunlight measured at 0.63 μm over oceans and has been corrected to total aerosol optical thickness at wavelength 0.5 μm. A "climatology" of the tropospheric aerosol loading is then subtracted from these data based on 1989 and 1990, two volcanically quiescent years. While this record is not a direct observation of stratospheric aerosol like SAGE II, it is a useful data set as a single-instrument record with good spatial coverage of the densest parts of the Pinatubo aerosol while the SAGE II/GloSSAC record depends on disparate sources of observations and SAGE II only covers the Earth very slowly.
To validate the model's treatment of gaseous SO 2 in the Pinatubo and Cerro Hudson plumes, we use data from the TOMS, the Television Infrared Observation Satellite (TIROS) Optical Vertical Sounder (TOVS) High Resolution Infrared Sounder sensor, and the Microwave Limb Sounder (MLS). TOMS is an ultraviolet sensor that takes measurements at 6 wavelengths while the High-Resolution Infrared Radiation Sounder Version 2 (TOVS/ HIRS/2) is an infrared sensor that observes 20 wavelengths. These two satellites are used here, as compiled by Guo et al. (2004), to estimate the total mass of the Pinatubo plume. In order to validate the altitude of the Pinatubo plume we use additional data from MLS from its first few operational days. MLS became operational on 19 September 1991 and was immediately used to observe the fading Pinatubo SO 2 plume. Finally, we use OPC observations from the long-term data set over Laramie, WY (Deshler et al., 2019). These balloon-borne measurements represent a rare in situ observation of stratospheric aerosol. They are useful in validating the output of a sectional aerosol model like CARMA independent of uncertainty added by the optical calculations required to compare with satellite observations. The series of balloon-borne particle counter measurements from Laramie from July to September 1991 (Deshler et al., 1992(Deshler et al., , 1993 demonstrate that the full depth of the Pinatubo cloud arrived above Laramie in September. Shallow layers of Pinatubo aerosol were measured from initial Laramie soundings in July (see Deshler et al., 1992), with a deeper layer during August, measured also from SAGE-II (Trepte et al., 1993) and ground-based lidar (e.g., Vaughan et al., 1994).

Results
GEOS CCM was used to simulate the impacts of volcanic aerosols from the 1991 eruptions of Mt. Pinatubo and Cerro Hudson on stratospheric chemistry and aerosols. We first evaluate the model in terms of the available satellite aerosol observations. Because SAGE II was saturated during the period immediately following the eruption, here we focus on the AVHRR record and the SAGE record supplemented by LIDAR observations during June 1991, as described by Thomason et al. (2018). According to AVHRR, the global average stratospheric aerosol optical depth (sAOD) reached a maximum of 0.155 3 months after the eruption, in September 1991 ( Figure 2). The 10 Tg injection ensemble more closely matches this behavior in the global average. All model results and the SAGE observations in Figure 2 have been masked to exclude latitudes without AVHRR observations each month (Figure 3e), which explains the sharp changes on the first of August and September at southern midlatitude (right panel). The ensemble using the 10 Tg SO 2 and the ensemble using the 20 Tg SO 2 Pinatubo injection reach a maximum of 0.16 on 1 November and 0.22 on 3 October, respectively. These two ensembles represent edge cases of estimates of the initial Pinatubo SO 2 injection and represent a range of possible Pinatubo-like eruptions. The AVHRR observations of the early Pinatubo plume are within this range. SAGE II observations of the early plume became saturated and were filled in lidar data from aircraft flights (Thomason et al., 2018). After the initial 3-4 months of increasing sAOD, both observations and simulations show declining sAOD as sulfate production slows and aerosols begin to be removed from the stratosphere.
The majority of the Pinatubo aerosol remains in the tropical stratosphere throughout 1991. AVHRR observed a maximum tropical (30°S-30°N) zonally averaged sAOD as high as 0.27 in September, shown in Figure 2. The SAGE II/GloSSAC observations reach a maximum in November at 0.16 zonally averaged tropical sAOD. The 10 Tg ensemble reached a maximum tropical zonally averaged sAOD of 0.16 on 3 October while the 20 Tg simulation reached a maximum of 0.26 on 1 September. The third panel of Figure 2 shows similarly calculated averages for the Southern Hemisphere midlatitudes (60°S-30°S) peak at 0.14 in the AVHRR data, 0.10 in the filled SAGE II data.
The comparison between the ensembles with and without the Cerro Hudson eruption indicates that about half of the sAOD in the Southern Hemisphere in 1991 was due to Cerro Hudson aerosols, which increased the sAOD by 0.1 over the simulation without Cerro Hudson. Note that both the SAGE II observations and the model have been masked to match the latitudinal extent of the AVHRR observations each month (see Figure 3e) before these averages were taken. The meridional spread, magnitude, and lifetime of the tropical Pinatubo aerosol in our simulations shows similar features to the observations from AVHRR. Figure 3 shows the zonally averaged sAOD over time in the simulations and observations. The latitudinal extent of the tropical Pinatubo plume in each of the simulations more closely follows the SAGE observations, remaining north of the Equator while AVHRR shows sAOD >0.3 as far South as 15°S. The impact of Cerro Hudson can be seen in both the AVHRR (Figure 3d) and SAGE II (Figure 3e) observations. The simulation without Cerro Hudson (Figure 3c) shows the contribution of the spread of Pinatubo aerosol south from the tropics, as in Figure 2c.
To evaluate the vertical extent of the Pinatubo plume in the model, Figure 4 shows the comparison of the GLOS-SAC observations of tropical (30°S-30°N) extinction coefficient alongside the model calculated extinction coefficient profile throughout the year. While there are differences in magnitude of the aerosol extinction between GLOSSAC and the modeled values, similar to in Figures 2 and 3, the altitude of the Pinatubo plume in the model is realistic, especially for the most optically dense part of the plume. The majority of aerosol extinction in the 20 Tg ensemble average is in a layer between 15 and 20 km in July; with self-lofting expanding the plume up to 25 km by the end of August. A similar pattern can be seen in the satellite observations-while the observed plume is almost entirely below 20 km in July, it reaches 28 km by the end of August and up to 30 km by the end   (Zhu et al., 2020). Airborne impactor collections on the NASA DC-8 surveys of the Pinatubo cloud (Pueschel et al., 1994) also show that the base of the cloud continued to contain volcanic ash particles up to 9 months after the eruption.
To evaluate the simulated quiescent and post-Pinatubo aerosol size distributions, we compare with balloon-borne OPC observations taken above Laramie, WY (41°N) in the months following the eruption (Deshler et al., 2019). Deshler et al. (2019) report observing an unperturbed stratospheric aerosol layer until 18 September 1991. Their next flight, on 2 October, showed a volcanically perturbed stratospheric aerosol. The observed aerosol count in each OPC channel increased significantly as the Pinatubo aerosol reached Laramie's latitude.
The exact timing of the arrival of the Pinatubo aerosol above Laramie varies between simulations and from the true transport because of the model's free-running meteorology. Figure 5 shows a comparison of the modeled cumulative aerosol size distribution compared with OPC measurements above Laramie, WY. In order to compare with the pre-and post-Pinatubo OPC data, a pre-Pinatubo size distribution is represented by a zonal mean at the latitude of Laramie (41.3°N) of the simulations in July and August and a post-Pinatubo size distribution is represented by a zonal mean of the simulations in November and December. Both the 10 Tg ensemble and 20 Tg simulation of the pre-Pinatubo stratospheric aerosol are within the spread of OPC observations taken between 17 June and 18 September. The simulations also both show an increase in number of particles by 15 November which generally match the observations taken between 2 October and 12 December. Both ensembles represent plausible increases in particle density due to the nucleation, growth, and transport of the Pinatubo plume into the Northern mid-latitudes, while the simulated pre-Pinatubo concentrations are in good agreement with the observations of the unperturbed stratosphere. Figure 6 shows a comparison of the simulations with a zonal average of MLS observations of the tropical (10°S-10°N) SO 2 on 21 September 1991 (Read et al., 1993). MLS became operational only on 19 September 1991, and so was not available to provide insight into the chemistry of the earlier part of the Pinatubo plume. The 21 September 1991 data provide a boundary condition for the plume chemistry up to that point. The means of the September monthly and zonal average of the 10 and 20 Tg ensemble members are calculated and compared with these observations. The magnitude of the zonally averaged SO 2 in the 20 Tg simulations is comparable to the MLS observations. Both the 10 Tg ensemble and 20 Tg simulation roughly match the height of the plume indicating a realistic rate of self-lofting.
Estimates of the integrated mass of SO 2 in the Pinatubo plume based on TOMS and TOVS HIRS were compiled by Guo et al. (2004) and are shown in Figure 7. These observations, similarly to the observations of sAOD in Figure 2, fit between the 10 Tg (red) and 20 Tg (blue) simulations. Note that TOMS and TOVS differ in their initial estimate of SO 2 mass by more than 1 Tg of sulfur (S), indicating significant uncertainty in the actual mass of the Pinatubo injection.
Finally, we estimate the impact of OH depletion on the conversion of SO 2 to sulfate in the Pinatubo plume. Observations of the early Pinatubo plume were sparse, but the SO 2 lifetime in the plume was characterized from estimates of the initial SO 2 mass injected and tracking of the plume in the subsequent weeks. The e-folding time of SO 2 within the Pinatubo plume, that is, the time it takes for the mass of SO 2 to decrease to 1/e of its initial value, has thus been observationally constrained to be between 23 and 35 days (Bluth et al., 1992;Guo et al., 2004). This metric for reporting the oxidation rate of SO 2 in the plume assumes a near-constant concentration of OH, resulting in a single e-folding time, no matter the time horizon it is calculated for. When calculated consistently in our model this "average" e-folding time, or the time for the Pinatubo SO 2 to decrease by a factor of 1/e is 34 days in the 10 Tg ensemble and 36 days in the 20 Tg ensemble, consistent with earlier GEOS simulations that used prescribed oxidant fields, as in Aquila et al. (2012). Importantly, though, those earlier "uncoupled" simulations were found to have too rapid of SO 2 to sulfate immediately after the eruption.
The e-folding time estimates noted above are effective over a long period in the run, but they do not illustrate the rate of change of the SO 2 consumption in the plume due to depletion of the oxidants. The coupled aerosol-chemistry model used here allows us to assess the SO 2 lifetime at a finer timescale than previous versions of the model. In order to quantify the dynamic lifetime of SO 2 in the oxidant poor plume, we calculate the instantaneous e-folding time of SO 2 in our model, as in Mills et al. (2017). Instantaneous e-folding time is defined as the time it would take to reach 1/e of the current mass at the current rate of oxidation, calculated using daily averages  (Aquila et al., 2012;Read et al., 1993) and modeled zonal mean tropical profiles from the 10 Tg (blue) and 20 Tg (red) simulations with a 95% confidence interval. The MLS observations are an average of 95 individual SO 2 profiles and the error bars show the retrieval error.
in the model. Figure 8 shows the instantaneous e-folding time for the 10 Tg ensemble and 20 Tg ensemble. In the 10 Tg ensemble, a maximum e-folding time of 169 days is reached the day after the eruption, before approaching an asymptote of 31.7 days by 1 August. The 20 Tg ensemble showed a more extreme OH response, reaching a maximum e-folding time of 202 days the day after the eruption before approaching an asymptote of 30.2 days by 1 August.

Conclusions and Discussion
We have used remote sensing data (SAGE II, AVHRR, TOMS, MLS, and TOVS) alongside in situ OPC observations to evaluate a new capability of the GEOS CCM to perform coupled aerosol, chemistry, radiation, and dynamics simulations. We focus here on the 1991 volcanic eruption of Mt. Pinatubo because it is much studied in the literature and gives us a benchmark to compare against, both in terms of observations and previous  (Guo et al., 2004). Similarly to aerosol optical depth in Figure 2, the 10 and 20 Tg case bracket the observed Pinatubo plume. Both cases have been fitted to an exponential decay function, the 10 Tg case asymptotes to 31.7 days while the 20 Tg case asymptotes to 30.2 days. models. Our model includes the detailed stratospheric and tropospheric chemistry of the GEOS-Chem mechanism coupled to the CARMA sectional aerosol microphysics model. This approach allows us for the first time to simulate with GEOS CCM the detailed microphysical response of volcanic sulfates and their perturbation to the quiescent stratospheric background aerosol in a framework that allows for realistic depletion of oxidants in the dense volcanic plume.
GEOS CCM simulates an eruption based on the 1991 eruption of Mt. Pinatubo. It simulates a perturbation to the background aerosol with similar magnitude and temporal evolution to what was observed in 1991. This work highlights this model's sulfur chemistry mechanism, which, as part of the broader GEOS-Chem mechanism, interactively simulates the impact of a large-magnitude explosive volcanic SO 2 plume on broader stratospheric chemistry. These simulations also show the model's skill in evolving a realistic aerosol size distribution following a volcanic eruption. The mass of injected SO 2 associated with the 15 June 1991, eruption of Pinatubo remains uncertain due to the lack of early observations and the coincident Typhoon Yunya. This work demonstrates that the global stratospheric impacts of Pinatubo can be modeled by an injection of SO 2 based on TOMS observations in the days following the eruption. Our model indicates that an injection of 10 Tg of SO 2 is adequate to explain space-based AOD observations but that early MLS measurements of SO 2 and in situ OPC observations may indicate that an injection of 20 Tg results in the model more closely matching observations in the final months of 1991.
We additionally use the model to characterize the impact of the Pinatubo and Cerro Hudson eruptions on Southern Hemisphere sAOD in the months following Pinatubo. According to our model, about 50% of the AOD in the Southern Hemisphere in the 6 months following Pinatubo can be attributed to Cerro Hudson. As shown in Figure 2c, our ensemble without the Cerro Hudson eruption produces an increase in Southern Hemisphere sAOD as Pinatubo aerosol are transported at high altitudes into the Southern Hemisphere midlatitudes, consistent with the findings of Aquila et al. (2013). In comparison, the simulations which included both eruptions show an additional 0.1 sAOD from the Cerro Hudson eruption. Our simulations indicate that including Cerro Hudson is important in reproducing observations of stratospheric aerosols in 1991.
These simulations also support the hypothesis that major OH depletion happened in the early Pinatubo plume. A decrease in the decay rate of SO 2 of similar magnitude to Mills et al. (2017) is calculated by the GEOS-Chem chemistry mechanism, represented by an increase in the e-folding time from a background value of 30-31 days to more than 200 days. While observations of SO 2 from the weeks following the eruption are too coarse to see this effect, our simulations indicate that this effect is significant in eruptions of similar magnitude as the 1991 eruption of Mt. Pinatubo. Further work is necessary to quantify the magnitude of this effect in smaller eruptions in the satellite record or in larger historical eruptions, like in Clyne et al. (2021). Zhu et al. (2020) also reported in a simulation of Kelut that SO 2 may undergo heterogeneous reactions on ash while Abdelkader et al. (2023) suggest that injected water vapor can enhance stratospheric OH. Such effects are not included in our simulations, but they might have resulted in a loss of SO 2 more rapid than in our simulations.
Further analysis of the 1991 eruption of Mt. Pinatubo is needed to fully understand the processes that led to its ultimate impacts on the Earth system. Specifically, our model, because of its resolution and global scope, is not equipped to resolve the meteorological situation on the day of the eruption. For this reason, we have not attempted to find the "true" mass of injected SO 2 from the eruption and instead have mapped the model's sensitivity to a roughly Pinatubo-sized eruption. A regional, finer resolution model would be better equipped to estimate the initial mass of SO 2 and other eruptive gasses and aerosols, as in G. Stenchikov et al. (2021). Even within the scope of global modeling, a better understanding of the interaction of SO 2 , sulfate aerosols, and ash aerosols in volcanic plumes could constrain the appropriate injection parameters and impacts of the Pinatubo eruption.
The model presented in this work could be used to characterize the dynamic stratospheric aerosol in the more recent volcanic record, and could inform observations of the stratosphere moving forward. The stratospheric aerosols satellite record after Pinatubo includes a volcanically quiescent period until 2000, followed by a decade characterized by many relatively small tropical eruptions, which have been shown to have caused the majority of sAOD variability (Carn et al., 2016;Neely et al., 2013). The model developed for this study is well equipped to characterize the stratospheric aerosol of this period. Volcanic ash aerosols, though generally short-lived, have been shown to alter the long-term impacts of volcanic eruptions via heterogeneous oxidation of SO 2 as well as acting as a sink for sulfate mass as they settle out of the stratosphere. Future model developments are planned to include these effects. Recent work has also shown the impact of aerosols from pyrocumulonimbus events on the stratosphere can be of similar magnitude to volcanic eruptions. We plan to incorporate carbonaceous aerosols into the model to study these events and their interactions with the background stratosphere and volcanic eruptions.